Vals AI Updates

Introducing the RSI Index: can a model do the research that builds the next model?

ProofBench v1.1 Released

DeepSeek's V4 Pro 0813 evaluated across our benchmark suite

Anthropic's Claude Opus 5 evaluated across our benchmark suite

Introducing the Web Search Index: native provider search vs. independent web-search tools

Google's Gemini 3.5 Flash Lite evaluated on the Vals Index

Google's Gemini 3.6 Flash evaluated on the Vals Index

Moonshot AI's Kimi K3 evaluated on the Vals Index

Thinking Machines' Inkling evaluated on the Vals Index

Meta's Muse Spark 1.1 evaluated across our benchmark suite